Protein synthesis is quantitatively determined by translation initiation and elongation.However, these two aspects have not been decoupled at genomewide scale.In this study, we used the information of translating mRNA as a key node to decouple these two aspects.We discovered that over 85% of the protein abundance can be quantitatively predicted by translating mRNA abundance and mRNA length in a multivariate manner, representing a quantitative rule in the Central Dogma in HeLa cells.We further characterized genome-wide translation initiation (translation ratio, TR) and elongation velocity and revealed two polarized modes of translational control.Per knowledgebase oriented functional analysis, the genes with high TR are focused on translation regulation and cellular proliferation, while those with low translation elongation velocities are characterized by distinct codon preference, remarkable translational pausing sites, low degradation rates and cancer-favorable features.In sum, we discovered a novel rule of translation control by either productivity-focused translation preference or quality-focused translation slow-down for anti-degradation on the genes that are essential for the maintenance of cellular phenotypes.